Sei sulla pagina 1di 11

Distinguished Lecturer Online (DLO)

Sponsored by

Society of Petroleum Engineers


Distinguished Lecturer Program
www.spe.org/dl 1

Interpreting Data from


Permanent Downhole Gauges

Roland N. Horne
Stanford University
July 17, 2013

Permanent Downhole Gauges


• More than 20,000 installed
worldwide.

• Usually installed to monitor


downhole equipment.

• Data rarely applied for


reservoir analysis.

1
Reservoir Engineering Uses

1. Reservoir pressure
2. Flowing bottomhole pressure management
3. Replacement for shut-in tests
4. Skin determination
5. Monitoring interference effects
6. Voidage control
7. Tubing hydraulics matching
8. Inflow performance modeling
9. Monitoring well treatments
10. History matching

4
Kragas, Turnbull and Francis (2004)

Replacement of Shut-In Tests

Northstar, Alaska
6 wells, 2 days duration
10,000 STB/d production would be lost per well

120,000 STB acceleration per campaign


650,000 STB acceleration over field life

And, $1.6 million avoided wireline costs

5
Kragas, Turnbull and Francis (2004)

PDG-Specific Issues

1. Manipulation and processing of large


numbers of data, including the filtering of noise
and nonidealities.
2. Deconvolution of the complex time series
into characteristic behaviors.
3. Identification of break points, to separate
transients into relevant subsections.
4. Accommodation of changes in well or
reservoir parameters (such as permeability
and skin) as a function of time.
5. Handling of flow rate information.
6. Temperature measurements.
6

2
1. Manipulation & Processing of Data

Data at 1 second frequency =


32 million data/year per gauge.
Physical storage and access are a
challenge for today’s databases.
Access, retrieval and transfer are a
challenge for today’s bandwidths.

7
Chorneyko (2006)

Permanent Gauge Data

6000 a
a
a
a
a
a
a
a
a a
a
a
a Outliers
aaa
5900 a

a Behavioral aa
aa
a
aaa
aaa a
a
a
a
a
a
a aa a
a a a
a aa

aberration a a a
Pressure (psia)

a a a

Noise
a a a
a
aa a a

5800 aa
a
aaa
aa
a
a
a
a
a
a
aa a
aa a
aaa a aa
aa
a aa
a
a a aaa
a aaa
a aaa
a a aa
a
a
a aa aaaaaaaaaaaaaaaaaaaaaaaaa a a
aaa
a aa a aaa
a aaaaaaaaa a a a a a a a
a

5700 a aa a a
a
a a
a a a
aa
a a a
a
a a a a a
a
a a aa
a a a a
aa a a a aaa aaaaaaaaaa a a a a a
a a a a
a
aaaaaaa aa a
a
a
a a a a aaa aa a
aaaa a a a
aaaaa a
a a
aa a
a a aaaaaaaa a a a a a a a aa a a
aaa a
a a a a
a
a aa a
aaa aa
a a
aaaaaaa a a a a a
aa a a
a a aa
a aa aaaaa

5600
a a
a aaa a a
a
aa a
a a aaaa
aa

Missing data
a aaaaaa
aaaaaa
a

5500
0 100 200 300 400 500
Ti me (hour)
8
(Athichanagorn et al., 2002)

Denoising with Wavelets

9
(Athichanagorn et al., 2002)

3
Outlier Filtering with Wavelets

10
(Athichanagorn et al., 2002)

2. Deconvolution
t
∆p w (t ) = ∫ q' (τ ).∆p0 (t − τ ) dτ
0
Theoretician’s playground, since 1949.
Remained impractical until recently.
Work of von Schroeter, Hollaender and
Gringarten (2004), using nonparametric
regression, p and q matching, derivative
restrictions and smoothness limit constraints.

11

2. Deconvolution
2820 10000
45 hours of data, no transient longer than 5 hours
2800
8000
Flow rate, bbl/day

2780
Pressure, psi

events 6000
2760
4000
2740

2000
2720

2700 0
0 5 10 15 20 25 30 35 40 45
Time, hrs
12
Nomura (2006)

4
2. Deconvolution
100

TRUE
case1
case2
Pressure derivative, psi

case3
45 hour response, inferred from deconvolution
10

1
0.001 0.01 0.1 1 10 100
Time, hrs
13
Nomura (2006)

2. Deconvolution – Issues

Model may change over time.


Buildups and drawdowns may be
different.
Levitan (2005): shut-ins only
Olsen & Nordvedt (2006): shut-ins only
Strong dependence on break points.

14

3. Transient Identification – Break Points


Wavelet approach: Athichanagorn et al. (2002)
2810
2800
2790
2780
Pressure, psi

2770
2760
2750
2740
2730
2720
0 5 10 15 20 25 30 35 40 45
15
Time, hrs Nomura (2006)

5
3. Break Points and Deconvolution
100
Wavelet
Adjusted
TRUE
Pressure derivative

Inaccurate break points are fatal for deconvolution

10

1
0.001 0.01 0.1 1 10 100
Time, hrs
16
Nomura (2006)

3. Break Points and Rate Normalization

17
Houzé (2006)

3. Break Points – Other Approaches

Wavelets often used for ‘first round’.


Khong (2001): statistical method.
Rai (2005):
Savitzky-Golay smoothing filter
Segmentation method
Olsen & Nordvedt (2005): pattern
matching method.

18

6
3. Break Points – Insertion/Deletion
case threshold initial insertion deletion
1 0.5 18 136 95
2 0.01 36 167 81
3 2 11 151 101
100

TRUE
case1
case2
Pressure derivative, psi

case3

10

1
0 .0 0 1 0 .0 1 0 .1 1 10 100
T im e , h rs 19
Nomura (2006)

4. k and s Changes in Time


• Constant property solution
• k(t)=constant , s(t)=constant

20
(Lee, 2003)

4. k and s Changes in Time


• Quadratic function for permeability
• k(t)= a(t-tL)2+kL , s(t)=s0 + b t

21
(Lee, 2003)

7
4. k and s Changes – Field Observations

Khong (2001)
Athichangorn, Horne and Kikani (2002)
Richardson, Roux, Quinn, Harker and Sides
(2002)
Lee (2003)
Haddad, Proano and Patel (2004)
Coludrovich, McFadden, Palke, Roberts and
Robson (2004)
Chorneyko (2006)
Olsen and Nordtvedt (2006)
22

4. k and s Changes

de Oliviera and Kato (2004): analytical


models used traditionally for
conventional well test interpretation may
be too simple to define the pressure and
flow rate transients that occur during the
extended duration of a permanent
downhole gauge record.
Using full-scale numerical models is
probably what we need, but not widely
applied (yet).
23

5. Downhole Flow Rate Gauges

Both p and q contain measurement errors.


Match both simultaneously.

1 r r2 nc r r r
obj = d − p + λ ∑ Rr2 (b ) + µ Da − q
2

2 r =1

Von Schroeter, Hollaender and Gringarten


(2004)
Nomura (2006)
24

8
5. Downhole Flow Rate Gauges

8910 8600
measured
8908 8400 iteration 1
Pressure (psia)

iteration 2
8200 iteration 3

Rate (STB/d)
8906
iteration 4
8000
8904 Liquid rate [bbl/d]
Pressure [psia]

7800
8902
7600
8900
7400
measured
8898
iteration 1 7200
iteration 2
8896 iteration 3 7000
iteration 4
8894 6800
450 500 550 600 650 450 500 550 600 650

Time (seconds)
Time (seconds)
Time (seconds)
Time (seconds)

25
Ahn (2008)

5. Downhole Flow Rate Gauges

A B

p
q

A – flow event B – noise event


26

6. Temperature Data

Duru
27
(2008)

9
6. Temperature Data

Match temperature
history porosity

28
Duru and Horne (2008)

Conclusions (1)
Permanent downhole gauges are rich sources
of reservoir data.
Not just more, but better!
Good progress on:
Noise and outlier removal
Break point identification
Deconvolution
Combining rate data
Utilizing temperature data
But, more work to do!
29

Conclusions (2)

The ultimate goal is to achieve a high


degree of automation.
Nobody wants to look at 100 million data
points!
Eventual inclusion in SmartFields
procedures.

30

10
Acknowledgements

Members of the SUPRI-D research


consortium on innovation in reservoir
testing.
SUPRI-D graduates:
Athichanagorn (1999)
Khong (2001)
Lee (2003)
Rai (2005)
Nomura (2006)
31

11

Potrebbero piacerti anche